(Kac–Moody) Chevalley groups and Lie algebras …...(Kac–Moody) Chevalley groups and Lie...

(Kac–Moody) Chevalley groups andLie algebras with built–in

structure constantsLecture 3

Lisa Carbone, Rutgers Universitylisa.carbone@rutgers.edu

July 18, 2017

TOPICS

(1) Overview and introductory comments

(2) Lie algebras: finite and infinite dimensional

(3) Weights, representations and universal enveloping algebra

(4) (Kac–Moody) Chevalley groups

(5) (Kac–Moody) Chevalley groups over Z, generators and definingrelations for (Kac–Moody) Chevalley groups

(6) Structure constants for Kac–Moody algebras and Chevalleygroups

Today we will answer the question of constructing (Kac–Moody) Chevalleygroups over Z and of associating defining relations to (Kac–Moody)Chevalley groups.

LAST TIME: (KAC–MOODY) CHEVALLEY GROUPS

Let g be a symmetrizable Lie algebra or Kac–Moody algebra over C.

Let K be an arbitrary field. Let αi, i ∈ I, be the simple roots and ei, fi thegenerators of g.

Let VλK be a K–form of an integrable highest weight module Vλ for g,

corresponding to dominant integral weight λ and defining representationρ : g→ End(Vλ

K).

For s, t ∈ K, let

χαi(s) = exp(ρ(sei)), χ−αi(t) = exp(ρ(tfi)).

ThenGVλ(K) = 〈χαi(s), χ−αi(t) | s, t ∈ K〉 ≤ Aut(Vλ

K)

is a simply connected (Kac–Moody) Chevalley group corresponding to g.

CONSTRUCTING A SIMPLY CONNECTED

(KAC–MOODY) CHEVALLEY GROUP

Recall that we chose a lattice Q ≤ Λ ≤ P between the root lattice Qand weight lattice P, which can be realized as the lattice of weights ofa suitable representation V.

The simply connected group has desirable properties when wechoose a highest weight module whose set of weights contains all thefundamental weights.

If we choose Λ = Q then GV is the adjoint Chevalley groupIf we choose Λ = P then GV is the simply connected Chevalley group

If Q = P, then a representation whose set of weights contains all thefundamental weights is V = Vω1+···+ω` , the highest weight modulecorresponding to the sum of the fundamental weights.

THE CHEVALLEY GROUP FOR sl3(C)Let sl3(C) denote the Lie algebra of 3×3 matrices of trace 0 over C.Let α1, α2 denote the simple roots. The basis of the standard 3dimensional representation ρ : sl3(C)→ End(V) of sl3(C) onV = C⊕ C⊕ C is:

xα1 =

0 1 00 0 00 0 0

, xα2 =

0 0 00 0 10 0 0

,

hα1 =

1 0 00 −1 00 0 0

, hα2 =

0 0 00 1 00 0 −1

,

x−α1 =

0 0 01 0 00 0 0

, x−α2 =

0 0 00 0 00 1 0

.

Here we are using the Chevalley notation where xαi = ei, x−αi = fi.The weights of ρ are:

ω1, ω2 − ω1, −ω2.

The representation ρ coincides with the highest weight module Vω1

with highest weight ω1.

LIE ALGEBRA sl3(C) TO CHEVALLEY GROUP SL3(C)

Let V = Vω1 be the highest weight module for sl3(C). Letρ : sl3(C)→ End(Vω1) be the defining representation. Let s, t ∈ C. InAut(Vω1), as before, we set

χαi(s) = exp(ρ(sxαi)), χ−αi(t) = exp(ρ(tx−αi))

But x2α1

= x2−α1

= 0 and x2α2

= x2−α2

= 0 thus

χα1 (s) = Id + ρ(sxα1 ) =

1 t 00 1 00 0 1

, χ−α1 (t) = Id + ρ(tx−α1 ) =

1 0 0t 1 00 0 1

,

χα2 (s) = Id + ρ(sxα2 ) =

1 0 00 1 t0 0 1

, χ−α2 (t) = Id + ρ(tx−α2 ) =

1 0 00 1 00 t 1

.

The lattice generated by the weights of Vω1 is the weight lattice P.Thus the Chevalley group GVω1 is simply connected.

SIMPLY CONNECTED CHEVALLEY GROUP

CORRESPONDING TO sl3(C)

The simply connected Chevalley group is the group GVω1 ≤ Aut(Vω1),generated by the automorphisms χ±αi :

GVω1(C) = 〈χαi (s), χ−αi (t) | i = 1, 2, s, t ∈ C〉

= 〈exp(ρ(sxαi )), exp(ρ(tx−αi )) | i = 1, 2, s, t ∈ C〉

=

⟨1 t 00 1 00 0 1

,

1 0 0t 1 00 0 1

,

1 0 00 1 t0 0 1

,

1 0 00 1 00 t 1

| s, t ∈ C⟩

This is the simple Lie group SL3(C).

ARITHMETIC SUBGROUP GV(Z)The arithmetic subgroup SLn(Z) of SLn(C) is obtained by takingZ–entries in the matrix representation of SLn(C).

This corresponds to taking ‘Z–points’

GV(Z) = 〈χαi(s), χ−αi(t) | s, t ∈ Z, i ∈ I〉

of the Chevalley group GV(C).

For GV(C) = SL2(C), this is the subgroup GV(Z) = SL2(Z) generatedby the matrices (

1 s0 1

),

(1 0t 1

)for s, t ∈ Z.

This is well known, but does not generalize to exceptional groups orto Kac–Moody groups.

A crucial fact for generalizing this construction is the following.

The subgroup SL2(Z) of SL2(C) is also the stabilizer of a Z–form VZ of thestandard representation VC of the Lie algebra sl2(C).

HIDDEN STRUCTURE

Proposition The subgroup SL2(Z) of SL2(C) is the stabilizer of a Z-form VZ of thestandard representation VC of the Lie algebra sl2(C).

Proof: Take VZ = Z⊕ Z and VC = C⊕ C. Then SL2(Q) acts on VC:(a bc d

)·(

xy

)=

(ax + bycx + dy

).

Suppose now that(

a bc d

)stabilizes VZ:

(a bc d

)· VZ ⊆ VZ, that is

(a bc d

)·(

xy

)=

(uv

),

where ad− bc = 1, x, y ∈ Z and u = ax + by ∈ Z, v = cx + dy ∈ Z.

Take(

xy

)=

(01

). Then u = ax + by implies b ∈ Z and v = cx + dy implies d ∈ Z.

Take(

xy

)=

(10

). Then u = ax + by implies a ∈ Z and v = cx + dy implies c ∈ Z.

Thus if(

a bc d

)· VZ ⊆ VZ then

(a bc d

)∈ SL2(Z). �

‘ARITHMETIC SUBGROUP’ OF A (KAC–MOODY)CHEVALLEY GROUP

For finite dimensional Chevalley groups, Chevalley defined thearithmetic subgroup GVλ(Z) as follows:

GVλ(Z) = {g ∈ GVλ(C) | g(VZ) = VZ} ≤ Aut(VZ).

This is the subgroup of GVλ(C) preserving the lattice VλZ in the

representation space Vλ.How does this compare with the ‘group of Z–points’

GZ = 〈χαi(s), χ−αi(t) | s, t ∈ Z, i ∈ I〉

ofGC = 〈χαi(s), χ−αi(t) | s, t ∈ C, i ∈ I〉?

When g is finite dimensional, it is straightforward to prove that

GVλ(Z) ∼= GZ.

When g is an infinite dimensional Kac–Moody algebra, this is a difficult andsubstantial theorem, proven by C-Liu.

SUBTLE ISSUES

If g =

(a bc d

)∈ SL2(Z) is written in terms of the generators

(a bc d

)= χ±α(t1)χ±α(t2) . . . χ±α(tk)

then it is not necessarily the case that the scalars ti are all integers:(1 11 2

)= χα(

12

)hα(12

)χ−α(12

) =

(1 1

20 1

)( 12 00 2

)(1 012 1

).

However, since g ∈ SL2(Z) and the χ±α(t) generate SL2(Z) for t ∈ Z,there exist integers s1, . . . , sn such that

g = χ±α(s1)χ±α(s2) . . . χ±α(sn).

THE CHEVALLEY GROUP E7(Z)Hull and Townsend, following Cremmer and Julia, discovered thefollowing form of E7(Z):

E7(+7)(Z) = E7(+7)(R) ∩ Sp(56,Z)

in the framework of type II superstring theory. Soule gave a rigorousmathematical proof that the E7(+7)(Z) of Hull and Townsendcoincides with the Chevalley Z–form of G = E7 given by

E7(Z) = {g ∈ E7(C) | g(VZ) = VZ} ≤ Aut(VZ).

Here VZ is the stabilizer of the standard lattice in the unique56–dimensional fundamental representation of E7.

GENERATING SETS

Theorem ([CL]) Let R be a commutative ring with 1. Let λ be adominant integral weight and let Vλ be the corresponding integrablehighest weight module with simply connected Kac–MoodyChevalley group

GVλ(R) = 〈exp(ρ(sei)), exp(ρ(tfi)) | s, t ∈ R〉

Let s, t ∈ R, u ∈ R× and set

χαi(s) = exp(ρ(sei)), χ−αi(t) = exp(ρ(tfi)),

wαi(u) = χαi(u)χ−αi(−u−1)χαi(u), hαi(u) = wαi(u)wαi(1)−1.

Then GVλ(R) has the following generating sets:(1) χαi(s) and χ−αi(t),and(2) χαi(s) and wαi(1) = χαi(1)χ−αi(−1)χαi(1).

GENERATING SETS FOR SL2(Z)

The simply connected Chevalley group SL2(Z) has the followinggenerating sets

(1) χα(1) and χ−α(1), corresponding to matrices

(1 10 1

)and

(1 01 1

).

(2) χα(1) and wα(1) = χα(1)χ−α(−1)χα(1), corresponding tomatrices

(1 10 1

)and

(0 1−1 0

)where s ∈ Z.

GENERATORS AND RELATIONS FOR CHEVALLEY

GROUPSSteinberg gave a defining presentation for finite dimensionalChevalley groups over commutative rings R, using the generatingsets that we have described.Tits gave generators and relations for Kac–Moody groups,generalizing the Steinberg presentation.In the finite dimensional case, there is a Chevalley type commutationrelation of the form

(χα(u), χβ(v)) =∏m,n

χmα+nβ(Cmnαβumvn)

between every pair of elements χα, χβ .Here u, v ∈ R, Cmnαβ are integers and the χα are viewed as formalsymbols in

Uα = {χα(u) | α ∈ ∆, u ∈ R} ∼= (R,+).

However, in the infinite dimensional case, Tits’ presentation ofKac–Moody groups has infinitely many Chevalley commutationrelations.

DETERMINING TITS’ KAC–MOODY GROUP

PRESENTATION

In the infinite dimensional Kac–Moody case, Tits determined thatwhenever a pair of real roots is ‘prenilpotent’, then there is aChevalley commutation relation necessary for defining theKac–Moody group. In order to make Tits’ presentation complete, weneed to:Explicitly describe the infinite set of prenilpotent pairs of roots.This usually requires us to:Explicitly describe the infinite set of positive real roots.There is no guarantee that either of these tasks can be carried out inpractice.

PRENILPOTENT PAIRSLet (α, β) be a pair of real roots and let W denote the Weylgroup. Then (α, β) is called a prenilpotent pair, if there existw, w′ ∈W such that

wα, wβ ∈ ∆re+ and w′α, w′β ∈ ∆re

−.

A pair of roots {α, β} is prenilpotent if and only if α 6= −β and

(Z>0α+ Z>0β) ∩∆re+

is a finite set. For every prenilpotent pair of roots {α, β}, Titsdefined the Chevalley commutation relation

(χα(u), χβ(v)) =∏

mα+nβ∈(Z>0α+Z>0β)∩∆re+

χmα+nβ(Cmnαβumvn)

where u, v ∈ R and Cmnαβ are integers.

If g is finite dimensional, every pairs (α, β) of roots isprenilpotent.

THE TITS–STEINBERG PRESENTATIONLet R be a commutative ring with 1. Let g be a finite dimensionalsimple Lie algebra or symmetrizable Kac–Moody algebra.We may associate to g a group G over R, generated by the set ofsymbols {χα(u) | α ∈ ∆re,u ∈ R} satisfying relations (R1)–(R7) below.Let i, j ∈ I, u, v ∈ R and α, β ∈ ∆re.

(R1) χα(u + v) = χα(u)χα(v);

(R2) For each prenilpotent pair (α, β),

(χα(u), χβ(v)) =∏

m, n > 0mα+ nβ ∈ Zα⊕ Zβ

χmα+nβ(Cmnαβumvn)

where Cmnαβ are integers.

(R3) wiχα(u)w−1i = χwiα(ηα,iu),

(R4) hi(u)χα(v)hi(u)−1 = χα(vu〈α,α∨i 〉) for u ∈ R∗,

(R5) wihj(u)w−1i = hj(u)hi(u−aji ),

(R6) hi(uv) = hi(u)hi(v) for u, v ∈ R∗, and

(R7) (hi(u), hj(v)) = 1 for u, v ∈ R∗.

THE DIFFICULTY OF FINDING PRENILPOTENT PAIRS

For g = e10, work of Allcock has shown that the problem ofdetermining all prenilpotent pairs is tractable but impractical.Namely, Allcock showed that for the number of W–orbits ofprenilpotent pairs of real roots having inner product equal to kgrows at least as fast as (const)k7 as k→∞.For each such orbit, we then have to enumerate theprenilpotent pairs of real roots.We need a different approach.

SIMPLIFYING TITS’ PRESENTATION

Tits’ presentation is very redundant and can be reduced significantly.In joint work with D. Allcock, we obtained a simplification of Tits’presentation for Kac–Moody algebras that are simply laced andhyperbolic, giving a finite presentation for these groups over Z.A Dynkin diagram is simply laced if it consists only of single bondsbetween nodes.This class includes the group E10(Z), conjectured to be a discretesymmetry group of Type II superstring theory.Our result shows that in a simply laced hyperbolic Kac–Moodygroup, the Chevalley commutation relations arising from simple rootsimply the Chevalley commutation relations arising from all realroots.

t t t t t t t t tt

α1 α3

α2

α4 α5 α6 α7 α8 α9 α10

A FINITE PRESENTATION FOR G(Z)Let g be a simply laced, symmetrizable and hyperbolicKac–Moody algebra.Tits’ presentation can be reduced to the following finitepresentation for G(Z):Over R = Z, the generators Xi(u) are obtained from Xi = Xi(1)via Xi(u) = Xu

i .

Any i Each i = {1, . . . , `} i 6= j not adjacent i 6= j adjacent

S4i = 1 [S2

i ,Xi] = 1 SiSj = SjSi SiSjSi = SjSiSj

S2i SjS−2

i = S−1j

Si = XiSiXiS−1i Xi [Si,Xj] = 1 XiSjSi = SjSiXj

S2i XjS−2

i = X−1j

[Xi,Xj] = 1 [Xi,Xj] = SiXjS−1i

[Xi,SiXjS−1i ] = 1

Table: The defining relations for G(Z), G simply laced and hyperbolic

COMPARING KAC–MOODY GROUPS

We have been considering the following group constructions:

− Tits’ presentation, which defines a Kac–Moody group,− Our Kac–Moody Chevalley group GV(C), for some choice of V

We conjecture that over C, Tits’ group given by generators andrelations coincides with our Kac–Moody Chevalley group GV(C) forsome choice of highest weight module V.

This involves the functorial properties of Tits’ Kac–Moody group,which are not completely understood.

In the finite dimensional case, this isomorphism is well known.

THE FEINGOLD–FRENKEL HYPERBOLIC KAC–MOODY

ALGEBRA AE3

Consider the hyperbolic Kac–Moody algebra with generalized Cartanmatrix (

2 −1 0−1 2 −2

0 −2 2

)It has been conjectured for some time that this hyperbolic algebra may be analgebra of internal symmetries of Einstein’s gravitational equations.The Weyl group W is the (3, 2,∞)-triangle group:

W = 〈w1,w2,w3 | w2i = 1, (w2w3)3 = 1, (w1w3)2 = 1〉,

which is isomorphic to PGL2(Z).The Dynkin diagram

is not simply laced and hence does not satisfy the conditions to giverise to our finite presentation.In recent work with Scott Murray, we give a recursive method fordetermining the infinite set of prenilpotent pairs of real roots in AE3.